Simulation is an essential method for scheduling discrete manufacturing processes. Because of disturbances and/or inaccuracy of the simulation model, the difference between the predicted and the real process will increase with time. This paper reports on a simulation system which is able to reduce this difference by synchronization and adaptation of the simulation model while the real process is running. The basic principles of the process accompanying simulation and an experimental system are described.

A lot of scheduling problems of flexible manufacturing processes can be described as combinatorial optimization problems. The sequence optimization problem, which asks for an optimal sequence of products, is a well known example. Although many different scheduling techniques were developed in the past, the application of them is very difficult. In our opinion the optimization of scheduling only can be solved satisfactorily in combination with methods of simulation which guarantee the required high flexibility of the optimization system. We describe in this paper an approach for an open optimization system, where the optimization cycle consists of two independent parts - the simulation and the search algorithm. The easy replacement of these parts is the advantage of the open architecture. So it is possible to combine several simulation systems and optimization algorithms in the same system and to use it as a experimental environment for examining new optimization strategies. Some optimization algorithms were already examined e.g. Genetic Search Algorithms and Tabu Search. The practical test will be carried out in the field of electronic production.

The production of large-sized machines or facilities is fundamentally different from the usual manufacturing processes. Mostly it is a fixed-site production, dominated by typical assembly processes instead of linear operation sequences in the known flow shop or job shop problems. To describe these processes, several graph methods were developed in the past, i.e. AND/OR-graphs or modified Petri nets, but most of them are static methods and they are not suitable for scheduling tasks. This paper demonstrates, how Timed AND/OR-graphs can be modeled with a common Discrete Event Simulation (DES) system. The objective is a simulation-based scheduling system for assembly processes which is able to improve the prediction of due date keeping as well as to optimize the workflow of the assembly systems.

This research is motivated by the need to verify and implement a schedule in a real production environment, especially in precarious production environments. This paper presents a mixed integer program (MIP) with time constraints and analysis risk parameters for tool interruptions. With the assistance of the survival analysis, a safety value will be computed and included in the MIP to downscale the available capacity. To verify the quality and robustness of the MIP, it is necessary to simulate tool interruptions and to change assumed release dates of production-bound jobs which have different stochastic distributions. To simulate these instabilities a hybrid model has been created which combines a discrete event simulation with a MIP solver. Finally, the results of the various simulations are compared.

The goal of this paper is to find optimal release dates for products and different manufacturing locations in semiconductor industry. The underlying scheduling problem is solved by constraint programming (CP). Lange [9] describes, that the scheduling problem is too complex to solve it with mixed integer programming and that CP also deals with requirements that makes scheduling complex. All this is applicable in this paper and that is the reason why CP is used. Different objectives are observed and the goodness of the computed suboptimal release dates is tested by a discrete simulation model. For the tests a simplified model of a backend test facility is used. At least the results are compared with dispatching and simulation based methods. All this is done under the aspect of rolling horizon planning.

In this paper a constraint programming (CP) approach for calculating release dates for lots within a supply chain environment is investigated. The lot start times are verified by a simulation model using different dispatching rules focusing on tardiness. To test the presented CP approach a simple fab model is constructed. The fab model consists of parallel batch machines as well as work centers and single machines. The investigated objectives are tardiness, earliness and cycle time. Due to the high complexity decomposition methods for the CP approach are tested. The results from the CP method are lot start dates which are verified by a downstream simulation run. The results show that the presented CP approach could outperform the simulation for all objectives. The content of this paper could be used as a first investigation for new scheduling methods within a supply chain management.

Currently machines in a parallel work center in semiconductor manufactory are assumed uniform in terms of impact on yield for most logic to dispatch schedule this machine set. But in reality machines are different even though they are allowed for the same products. In some layer forming areas machines can get a so called health parameter which describes the current condition of the machine. A high health value means, that defects produced with a machine are less probable. Also the products, which are processed at this work center, differ in their complexity and wafer area used for one chip. The goal is to schedule products with a high complexity and a larger chip size to those machines with the best health value. Doing so will minimize defect wafer area. For this, different dispatching rules and a mixed integer programming approach are compared within a simulation model for practical test data.

A high yield is extremely important for the costs of production and thus also the price and competitiveness. In this paper a scheduling method is presented, which takes machine parameters into account. This is done exemplary for a SMD manufacturing. Here, not all machines within a parallel work centre are assumed to be equal, even if they are capable to process the same products. Typically, some machines perform better than other machines. This machine performance could be described by a so called machine health parameter. Also, some products could be more important than other products. This may be defined i.e. by the number of PCB layers. The goal is to schedule more important products to those machines with the best health value. Doing so will increase yield for these products. For this, different dispatching rules and a mixed integer programming approach are compared within a simulation model for practical test data.

Manufacturing is today often characterized by a growing number of customer-specific products that have to be manufactured and delivered in given lead times, according to concrete delivery dates. Thus, highly relevant questions like When to start a production order at latest, in order to stay within my lead time? are answered by more or less primitive, backward-oriented planning approaches and without taking into consideration uncertainty or alternatives. It gets more complex, if different products are to be produced and the more complex the underlying manufacturing system is (e.g. semiconductor with re-entry cycles). These questions could be answered more specifically, more detailed and more robust, if discrete, event-based simulation (DES) would be applied in a backward-oriented manner. This paper describes evaluation results from the semiconductor domain and names restrictions and limits. They also show, that the backward-oriented simulation approach can be applied successfully for the scheduling of customer-specific orders.

This paper presents an approach for scheduling different types of preventive maintenances (PMs) for a work center of a semiconductor manufacturing facility. The PM scheduling problem includes time-dependent synchronization constraints and is implemented in a constraint programming model. A mix of periodic and workload-specific maintenances is scheduled considering the synchronization to available engineers which have individual shift schedules and skills that define the range of feasible maintenances. This also comprises maintenances having process durations covering multiple shifts, which requires a continuous availability of sufficiently skilled engineers. Additionally to the PMs, also handling and maintaining of unscheduled downs is considered in the model. Multiple objectives are investigated and used for optimization and tested on realistic data. To compare the results an additional simulation model is built up.

Besides common key performance indicators mainly so called operating curves are used to assess the performance of manufacturing systems. The advantage of the operating curves is that one can study the system behaviour under varying conditions. However, the systems cannot be compared with each other so easily. To ensure the comparability of the system behaviour the operating curves must therefore be normalised. On the example of a simple simulation model a method will be introduced, to get normalised operating curves. This method is also suitable for more complex or real manufacturing systems.

Ensuring a high uptime for manufacturing machines is crucial for efficient and cost-effective production. In opposite, preventive maintenance tasks (PMs) are necessary to assure the reliability of manufacturing processes. This also prevents serious and unpredictable machine crashes, which affect uptime and process scheduling. However, PM tasks themselves lower the uptime and are to be smartly scheduled within their given domain considering some requirements as ensuring the availability of a sufficient number of qualified engineers for the concerning period. This work investigates a PM scheduling problem with time-dependent synchronization constraints for a lithography work center. For this, a constraint programming (CP) modeling approach including decomposition is used. Multiple objectives are considered. For example, the minimization of crew backup violations or scheduling PMs according to the work in process (WIP) for embedding upcoming PMs smoothly into the system work load. This minimizes negative effects on throughput and tardiness.

To assess the performance of a manufacturing system is not as easy as it seems at first. Besides common key performance indicators (KPI) mainly so called operating curves are used to this end. The advantage of the operating curves is that one can study the system behavior under varying conditions. However, the systems cannot be compared with each other so easily. To ensure the comparability of the system behavior the operating curves must therefore be normalized. In this paper a method will be developed for defining normalized operating curves, which is also suitable for complex manufacturing system structures. These methods are based on Little’s law as well as a special kind of flow graphs, with which it is possible to find the limiting bottlenecks in the system under special conditions. An example will explain that the method works.

Semiconductor frontend fabs are very complex manufacturing systems. Typically, the bottleneck of such a fab is the photolithography area because of its highly expensive equipment and the huge number of required secondary resources -- the so called reticles. A reticle (mask) is needed to structure different layers of integrated circuits on the wafers. The reticles can be moved between the equipment with regard to several constraints. This paper examines the benefits of a solver-based reticle allocation in comparison to a classical rule-based heuristic. In a first part, several simulation experiments are performed on the basis of representative test data. The second part presents results from real world application. Thereby it is shown, that the new approach shows significant improvements of different key performance indicators (KPIs).

The effort for scheduling real manufacturing systems is generally very high for mathematical as well as for simulation-based methods. Combining both methods is the key for solving complex scheduling problems. The paper introduces a special approach, where at first a static resource allocation problem is solved by mixed integer programming (MIP). Based on the resulting reduced dedication matrices, feasible schedules are then generated by a discrete event simulation (DES). Possible applications can be found in many parts of the semiconductor manufacturing process, for example in the wafer test. The investigated wafer test consists of two pronounced bottlenecks; each of it is formed as a workcenter with its own dedication matrix. After testing the method with practice oriented benchmarks, the benefits of the approach are shown on data derived directly from the semiconductor manufacturing process.

In modern semiconductor manufacturing and primarily in high-mix-low-volume facilities it is increasingly important to ensure throughput and machine utilization requirements are met while satisfying tight goals in object tardiness at the same time. This is especially a challenge for the field of wafer test with its natural fluctuations and uncertainties of test times. A further important objective is the lowering of the work in process (WIP) for the purposes of minimizing costs held in the system and improving production predictability. For this, the Virtual Time Based Flow Principle (VTBFP) -- a partly synchronized control strategy - is investigated in this paper. Tests are performed on a complex system, which is close to reality. As a result it is shown the benefits but also the limitations of the VTBFP approach.

One of the most promising approaches in modern microelectronics is the introduction of 3D chip micro systems with through-silicon via (TSV) interconnections. A successful transfer of this technology from the scientific level up to the level of mass production is not least of all a matter of cost-effectiveness and profit, which is directly related to high productivity. The developed technologies therefore have to be feasible for effective mass production. In this paper we introduce a method for planning and evaluating costs in future process chains. This method goes beyond usual mostly Excel-supported solutions, as it is based on a discrete event simulation system. The simulation model is generically generated out of an XML process chain definition file and includes a sophisticated state model for machines. Multiple process scenarios are created with the help of a supporting software tool. These scenarios are investigated for achieving favourable equipment and process chain configurations as well as control strategies to support manufacturing ramp-ups.

The following article is a case study for scheduling of assembly processes. Two optimisation methods, mixed integer programming solver and heuristic simulation-based optimisation are compared to each other. It can be shown, up to which model complexity the solver methods have the advantage over the simulation-based methods concerning a possible online application. Both the mathematical formulation as well as the discrete event simulation system, bases on a meta-model – a special Petri net. It is shown that bipartite graphs, i.e. Petri nets, avoid ambiguities in the case of alternative assembly sequences and have advantages in comparison to the usual simple precedence graphs. The results of optimisation experiments are described by the example of assembling a numeric controlled milling table. The model is capable of reflecting multiple parallel assembly processes, alternative assembly steps and can predict characteristics like utilisation, stock or adherence to schedules.

The optimization of manufacturing flows in electronics and semiconductor industry becomes more and more important. This results from the high complexity of the underlying production processes. In this research a selected part of the process with high practical impact is investigated. It is a scheduling problem arising in the semiconductor frontend oxidation and diffusion area. The optimization objective is the total weighted tardiness (TWT), but also the changes in cycle time are observed. Because of comparably long processing times and parallel batch processing a high optimization potential exists at the investigated machine groups. The methods dispatching, simulation-based optimization, mixed integer programming (MIP) and variable neighbourhood search (VNS) are compared. In contrast to the traditional simulation-based optimization, where a control variable is selected before a simulation run starts and is kept during the simulation, the new iterative simulation-based optimization approach is capable to generate and optimize small local sub-problems. Afterwards the results are used in a global model. The performance of the new method is comparable to MIP and VNS approaches and is much better than dispatching.